A locally segmented reconstruction method for parallel imaging.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Medical Imaging
• Signal Processing

Background:

• Parallel imaging reconstruction methods in MRI are crucial for accelerating image acquisition.
• Conventional methods can suffer from residual artifacts due to limitations in utilizing receiver coil sensitivity distributions.

Purpose of the Study:

• To propose and evaluate a locally segmented parallel imaging reconstruction method.
• To efficiently utilize the sensitivity distribution of multichannel receiver coils for improved image reconstruction.

Main Methods:

• Introduced a method of locally segmenting MR signals to maximize inter-channel sensitivity differences.
• Applied 1D Fourier transformation to undersampled k-space data, creating a hybrid 2D space.
• Partitioned the hybrid space into localized segments, estimated relationship kernels from autocalibration signals, and reconstructed images using sampled and estimated data.

Main Results:

• Computer simulations and in vivo experiments demonstrated fewer residual artifacts compared to conventional methods.
• The proposed method achieved performance gains by maximizing encoding capability of receiver channels.
• Accurate estimation of kernel weights reflecting adjacent signal relationships was achieved.

Conclusions:

• The proposed spatial segmentation method effectively reconstructs images with reduced artifacts.
• This approach maximally utilizes sensitivity differences among receiver channels for enhanced image quality.